Munich Personal RePEc Archive
The Size and Structure of Government
Michael, Bryane and Popov, Maja
University of Hong Kong, Government of Serbia
2011
Online at https://mpra.ub.uni-muenchen.de/53283/
MPRA Paper No. 53283, posted 01 Feb 2014 15:49 UTC
The Size and Structure of Government
Bryane Michael, Columbia University - SIPA
Maja Popov, General Secretariat of the Government of Serbia
Does government size and structure adapt to changes in government’s organisational environment (particularly to uncertainty and complexity) as predicted by organisational theory? We find – using a range of statistical analyses – support for each of the major theories of organisation adaptation (the contingency-based view, resource-based view, and rational choice view). We find that both government size and structure change – holding other factors constant – for changes in the uncertainty and complexity of governments’ organisational environments. We find seven clusters of governments which adapt their organisational sizes differently in response to changes in the
uncertainty and complexity of their organisational environments – and four clusters of governments with differing preferences for the way they adapt governmental structures.
We also use the available data to divide governments according to the extent to which they adapt their organisational size and structure reactively (after changes occur in their organisational environment), contemporaneously or strategically (before these changes in their organisational environment occur).
JEL Codes: F4, D7, E6, H1, H4
Keywords: contingency theory, public sector organisational theory, resource-based view, size of government, government structure
The views expressed in this paper remain the views of the authors alone and do not reflect the views of the organisations for which the authors work or are affiliated with.
Table of Contents
Introduction ... 3
What Do We Know About the Way the Size and Structure of Government Responds to Changes in the External Economic Environment?... 4
A Brief Background on the Size of Government ... 4
The Size of Governments Respond Weakly to a Changing Economic Environment ... 8
A Background on the Structure of Government ... 17
The Structure of Government Responds More to the Changes in the Policy Environment... 22
Literature Review... 27
Return of Contingency Theory: Government Size and External Shocks... 28
Organisational Buffering: Government Size and Domestic Output Fluctuations... 33
The Structure of Government: Composition of Expenditure and Decentralisation ... 37
The Public Sector Organisational Theory Literature ... 44
The Model ... 46
Empirical Results ... 52
Hypothesis 1: The size of government depends on the uncertainty and complexity of government’s organisational environment... 55
Hypothesis 2: Different governments will have different preferences for responding to the uncertainty and complexity of the macroeconomic environment ... 57
Hypothesis 3: Different countries will adapt reactively, strategically or contemporaneously to changes in the macroeconomic environment. ... 59
Hypothesis 4: Government structure changes in response to changing uncertainty and complexity of the macroeconomic environment... 61
Hypothesis 5: Differences in the uncertainty and complexity of various countries’ macroeconomic environments significantly explain changes in government structures. ... 64
Hypothesis 6: Different governments’ structures can respond strategically, contemporaneously or reactively to changes in their macroeconomic environments... 67
Conclusions ... 69
Appendix I: The Model ... 73
Appendix II: Data Sources and Quality ... 88
Appendix III: Empirical Analysis ... 92
The Size and Structure of Government Bryane Michael, Columbia University (SIPA)
Maja Popov, General Secretariat of the Government of Serbia Introduction
Despite over 40 years of theorizing about public sector organisation, we still know very little about how government responds – writ large – to changes in its organisational environment. A variety of theories predict how government size and structure should respond to the national macroeconomic environment it regulates and buys and sells labour, capital and goods in. Contingency theorists argue -- though are now in relative disrepute -- that government departments and agencies grow, shrink, divide and/or merge in response to changes in the macroeconomic environment (Gupta et al., 1994).
Resource theorists – and their newer off-spring who write about “competencies” -- argue that these government departments morph, depending on the resources (budgetary, staffing, know-how and so forth) they already have available – or can obtain through bureaucratic and/or political means (Bryson et al., 2007). Rational-choice theorists, and select scholars in public administration, argue that government organisational structure does (or should) foresee upcoming challenges and respond to them before they occur (Robertson et al., 1993 and especially Vietor, 2007). Finally, a new school of
interpretative and post-modern scholars argue that government organisational structure reflects cognitive understandings, culture, politics and symbols which no empirical study can correctly capture – or even try to (Frumkin and Galaskiewicz, 2004). Yet, despite these 40-plus years of studying public sector organisational theory, most primers about organisational theory in the public sector contain almost no actual empirical studies of the theories they present (Christensen et al., 2007).
Different governments’ size and structure responds to its organisational environment (particularly the uncertainty and complexity of that environment) differently. Some governments’ size and structure responds more to the resources they have at their disposal more than to changes in their organisational environments. In no case did we find that government size and/or structure does not correlate at all with changes in the uncertainty and complexity of the government’s organisational environment (as we define them in this paper) – militating against institutionalist theories of government (which argue that cognitive and internal processes drive the size and structure of government more than external factors). In this paper, we answer three questions. First, do different countries’ government size and structure respond to the uncertainty and complexity of the government’s organisational environment (as we have defined these terms in our paper)? Second, do such changes in organisational adaptation occur strategically (in the year before changes in the macroeconomic environment),
contemporaneous (in the same year as changes in the macroeconomic environment) or reactively (in the year after such changes)? Third, do some governments’ organisational changes correspond with changes in resources available rather than external changes in the government’s organisational environment?
While we arrive at many interesting (even exciting) results, the reader should keep three caveats in mind. First, we base our conclusions on very narrow proxies which we hope reflect the broad issues we study. For example, we base conclusions about changes in the structure of government on macroeconomic data about the composition of government expenditure – and hope that such changes reflect on the actual organisational changes we hope to explain. Second, we rely on statistical methods to find patterns in the data
(where sometimes such patterns may reflect over interpretation rather than actual fact).
We group countries using similarities in the way they spend government money and often treat statistical significance as practice significance. Third, we take 40 years of rich and deep research about public sector organisation – and reduce it to summary statistics based on data from the World Bank and IMF (often labelling governments’ patterns of organisational adaptation with politically controversial labels). Given these caveats, we most certainly do not wish to pass any final judgments on the rich, already-existing public sector organisational theory literature. Instead, we hope the empirical patterns in the data stimulate debate and encourage a new generation of scholars to take-up again empirical methods in the study of public sector organisation.
What Do We Know About the Way the Size and Structure of Government Responds to Changes in the External Economic Environment?
A Brief Background on the Size of Government
The sizes of governments around the world vary between about 10% of GDP to over 50% of GDP. A cursory glance at Figure 1a shows few similarities between countries which allow for generalisations about government sizes. Lesotho, the Maldives, Greece, Hungary and France have some of the largest governments – in terms of the amount of national resources managed and spent by the government (spending about twice the world average).1 Two countries often thought to be very different – Sweden and the USA – (on a world scale) have rather similar levels of government spending. Countries often noted for having relatively weak state capacity have some of smallest governments in the world (with the exception of China).
0%
20%
40%
60%
Lesotho Maldives
Greece Hun
gary Franc
e UK
Sweden Wo
rld USA Afghanistan
China Bang
ladesh Laos
Cent. Afri. Rep.
Cambodia
Figure 1a: Little Explains Differences in the Size of Government Around the World
Biggest 5 interesting m iddle countries Sm allest 5
Note: The data in the figure show government expenditure as a percent of GDP for 14 "interesting" countries (and a w orld average) out of a set of 124 countries for w hich the IMF provide data. In cases w here 2009 data w ere unavailable, w e used data for the latest year available since 2004.
Source: World Development Indicators (2010).
The data do not support the conventional view that government sizes correspond with citizens’ preference for public goods. In theory, the size of government should depend
1 We assume, like most authors writing about the size of government, that government expenditure as a percent of GDP serves as the most relevant indicator of such size. Other measures used in the literature include employment by the government (at various levels), levels of government consumption,
government revenue (earned through tax and non-tax methods). These other measures of government size correlate highly with government expenditure.
on the level of public goods and services demanded by a country’s citizens. As Swedes, Americans and Mexicans demand more roads, hospitals and other large goods which no one individual can pay for (or exclusively benefit from), government needs to collect and spend more.2 Figure 1b shows the correlation between citizens’ opinions about the importance of government in providing (goods, services and social protection) for all citizens and the size of their government. No relationship appears to exist between the proportion of GDP spent by government and the importance of government assigned by survey respondents among low-income, medium-income and high-income countries.
High-income countries tend to have larger governments and low-income countries tend to have smaller governments (judging by the few low-income countries for which the IMF provides data). Yet, the conclusion clearly emerges from these data that citizens’
preferences for public goods (and government writ large) do not seem to explain the size of government.
0%
10%
20%
30%
40%
50%
3 4 5 6 7 8
im portance of governm ent (higher m eans m ore im portant) size of government (higher numbers mean more important)
High Income Countries Low Income Countries Medium Income Countries Figure 1b: Attitudes about the Importance of Public Goods Offer Little Explanation
for Differences in Government Size Across Countries
Data in the figure show the importance given by survey respondents to the role of government and the proportion of government expenditure as a percent of GDP. Respondents ranked - betw een 1 and 10 - the extent to w hich they thought "government should take more responsibility to ensure that everyone is provided for." Government expenditure as a percent of GDP measures the size of governnment on the ordinate-axis. We use data for the latest year available since 2005.
Sources: World Development Indicators (2010) and World Values Survey (2010).
Government expenditure has grown almost everywhere in the world – mostly due to raising civil servant nominal salaries. Figure 2a shows the average change in
government expenditure for high-income and medium-income countries throughout the 2000s.3 In both groups of countries, government expenditure increased – though much more for medium-income countries than for high-income countries. Between 1999 and 2009, on average, high-income countries’ governments increased their expenditure (relative to GDP) by about 5%. Employment in these countries generally fell very slightly at the central level and rose very slightly at the local level – as well as in
government bodies like the social security administration and in state-owned enterprises.
Across various levels of government, expenditure by medium-income governments increased much more – by about 10%. In these medium-income countries, employment also remained stagnant; while nominal compensation to (government) employees increased by about 15% or more. As such, the story of expanding government – for
2 A number of non-economic explanations – like citizens’ desire to use government programmes to ensure justice in society or promote good citizenship – help explain differences in government size. We do not discuss (or analyse) these factors as these values and preferences change more slowly over time and prove more difficult to study using economic methods.
3 We use the terms medium-income and middle-income countries interchangeably throughout the paper.
These terms refer to the World Bank’s classification of countries by levels of income-per-capita.
high-income but especially for medium-income countries -- lies in paying existing workers more (in nominal terms) rather than hiring more staff.
-5%
5%
15%
25%
35%
Central Local General Central Local General
rates of change (in percent)
Size (expenditure) Employment Wage
Figure 2a: Civil Servant Nominal Wages Expand in the 2000s - More in the Medium- Income Countries that in the Rich Ones
High-Incom e Countries Medium -Incom e Countries
change in expenditure change in
employment
change in w ages paid
The figure show s w eighted averages of rates of change of overall government expenditure, nominal w ages and employment in countries for w hich the IMF and ILO provide data. For each individual country, w e calculate these rates of change as the simple arithematic average betw een 1999 and 2009 of the annual grow th rates of each of these three variables. We obtained w eights for government expenditure by taking the country's US dollar government expenditure in 2007 and divided such expenditure by the total US dollar government expenditure for the country group w hich the country belongs to (either high-income or medium-income). We obtained w eights for changes in employment by finding each country's proportion of employment as a percent of the total number of people employeed in general government for the income group the country belongs to in 2007 (or for the nearest year for w hich data are available). For changes in the compensation of employees (w hich proxies changes in w ages), w e w eight rates of change by 2007 dollar government expenditure (as the IMF provides data for such compensation in local currency and thus w e can not simply add pay across countries). Both the IMF and ILO provide data at the central, local and general government levels -- allow ing us to provide averages across these levels of govenment. The IMF and ILO do not provide data on enough low -income countries in order to draw valid averages for low income countries in general.
Sources: World Development Sources (2010) and ILO Employment database (2010).
The growth of capital in the public sector does not really explain the expansion of government sizes in the 2000s. Figure 2b shows the average change in financial and non-financial assets (the nearest proxy to capital one can obtain using public data). High income countries loaded-up on financial assets during the period (almost doubling the amount of financial assets they held) and generally divested from non-financial assets.
Medium-income countries tended to do the reverse – slightly divesting from financial assets and focused on acquiring non-financial assets. The rapid acquisition of financial assets during the end of the decade – during the financial crisis – only partly explains the overall acquisition of financial assets by high-income countries during the period.
-50%
0%
50%
100%
150%
High-Income Countries Medium-Income Countries
Expenditure Financial assets Non-financial assets
Figure 2b: High Income Governments Accumulated Financial Assets Whereas their Poorer Cousins Accumulated Physical Ones
The data in the graph show the average change in expenditure and the net acquisiton of financial and non-financial assets betw een 2000 and 2009 for all countries for w hich data are available. Period averages represent the simple arithemathic average of rates of change for each country over the 9 year period. In order to arrive at the average rates of change for high-income and middle-income countries, w e used each country's current 2004 government expenditure (expressed in dollar terms) in order derive w eights. For example, Germany's government expenditure (expressed in dollar terms) comprised roughly 11% of the total expenditure for the high-income country group.
Thus, Germany's average rate of change of government expenditure contributed 11% of the total w eighted average value for the high- income country group. We used a similar procedure to find group averages for changes in financial and non-financial assets. We used expenditure w eights because the IMF reports financial and non-financial assets only in local currency. The IMF does not provide enough data to provide similar comparisons for low -income countries.
Source: IMF's Government Financial Statistics (2010).
Differences in the type and variability of economic shocks in these government’s
organisational environment (as we define it in this paper) may in-part explain differences in government sizes across countries.4 Figure 3a shows the variability of GDP over the period 2000-2008 for selected high-income, medium-income and low-income countries.
Low-income countries’ GDP varied more throughout the period than GDPs in the other income groups. The most volatile economies in the high-income countries had variances similar to the most volatile economies in the medium-income countries group. The least volatile economies in all three income groups exhibited very similar levels of
(non)volatility during the period – suggesting that income-level itself makes a poor predictor of the volatility (and thus uncertainty) of a national economic environment.
0 3 6 9
Italy Jap
an Germany
Denm ark
Poland Bermuda
Sao Tome Australia
Cyprus Bah
rain Cent. Afr. R
ep. Liberia
Haiti Guinea-Bissau
Cong o DR Madaga
scar Ugan
da Ben
in Tanzania
Lao PD R Bangladesh
Iraq Gabon
Seych elles Cote d'Ivoire
Venezuela Yemen,
Rep.
Cameroon China
Vietnam
standard dev. of GDP
Figure 3a: Different Countries Have Very Different Economic Environments
high-incom e econom ies m edium -incom e econom ies low -incom e econom ies
The data in the graph show the volatility of GDP betw een 1999 and 2009 for all countries for w hich the IMF provide data. We show examples of countries w ith the highest and low est GDP volatility in each income group. We measure such volatility as the standard deviation of the log of GDP over the period divided by the average log value of GDP during the period. We used the natural log of GDP (rather than GDP itself) because using log values removes the effect of relvative size (as bigger economies w ill have a larger volatility of GDP simply because of their size). Log values - by their nature - describe changes in magnitudes. Thus, by looking at the standard deviation of the log of GDP, w e are focusing on changes in the magnitudes of GDP over the period rather than levels themselves.
Source: World Development Indicators database (2010).
A more detailed analysis of asymmetric macroeconomic shocks reveals much about the uncertainty of various governments’ organisational environments. Figure 3b shows the magnitude and timing of asymmetric shocks (shocks which affect one sector of the economy rather than the entire economy) for high-income, middle-income and low- income countries. The figure specifically shows changes in output in the industrial sector (as a percent of GDP) relative to changes in the service sector and/or the agricultural sector. The index we show in the figure rises as more resources are drawn into the industrial sector – and falls as more resources are pulled into the service or agricultural sectors. All three income groups have roughly the same magnitude of changes in sectoral production – albeit at different times. High-income countries tended to have larger volatility (measured by changes in the change) in industrial output than countries in the other income classification groups. Medium-income countries tended to have more steady growth rates in industrial output (with far fewer swings in the value of industrial
4 The canonical definition of an organisational environment from the organizational theory literature defines such an environment as the “forces outside the boundaries [of the organization] that can impact upon it [the organization]” (Hatch, 2006). In this paper, we focus on the macroeconomic environment and leave out the other elements such as legal environment, societal, and other environmental factors in order to limit the scope of our analysis.
production). Low income countries tended – in general – to show much less inter- sectoral macroeconomic volatility than the simple measure of GDP volatility we used in Figure 3a above shows. For many low income countries, the size of GDP throughout the period varied much more than the composition of that GDP between the industrial, service and agricultural sectors. In all cases, the variance or change in the broader macroeconomic environment makes the government’s organisational environment more uncertain – as both government and businesspersons have greater difficulty deciding to which sector of the economy they should allocate resources.
-20%
-10%
0%
10%
20%
30%
40%
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
changes in industrial production (relative to other industries)
High Income Low Income Middle Income
Figure 3b: Different Profiles of Uncertainty in Governments' Organisational Environment
The figure show s the changes in the relative proportion of GDP in the industrial sector as opposed to in the service sector or the agricultural sector. We use these changes as a proxy w hich might show the effects of sector-specific, asymmetric shocks (and thus measure the overall uncertainty of the macroeconomic environment). We constructed our proxy as follow s. We subtracted the proportion of the service sector in overall GDP from the proportion of industry in overall GDP (giving the absolute change in the importance of the service and industrial sectors). We divided these differences by the proportion of GDP in the agricultural sector (thus expressiing all "shocks" relative to the size of the agricultural sector). We calculated the rates of change of these ratios for each year (removing any rates of change over 300% or -300% w hich might have popped up due to the country having a relatively small agricultural sector). We found the arithematic average of these grow th rates betw een 1999 and 2009 and calculated a w eighted average of these grow th rates for each of the three groups of countries (high-income, medium-income and low -income).
We used each country's share of 2004 GDP in current US dollars (as a proportion of the total GDP for that county's group) as the w eight applied in our w eighted average calculation.
Source: World Development Indicators (2010).
low -incom e countries' changes in inter-sectoral
output high-incom e countries'
changes in inter-sectoral output
m edium -incom e countries' changes in inter-sectoral output
The Size of Governments Respond Weakly to a Changing Economic Environment How does government size respond to changes in the uncertainty and complexity of the government’s organisational environment? In the previous section, we showed the varying degrees to which the organisational environment of various governments around the world changed during the 2000s. Variance in GDP represents a simple proxy for the uncertainty and complexity of governments’ organisational environment (and we will discuss more refined measured later in the paper). More volatility in GDP makes planning more difficult – thus increasing overall uncertainty. More volatility also likely corresponds with more complex economies – because more complex economies have a greater need to reallocate resources across economic sectors, respond quickly and effectively to changes in tastes and technologies – and so forth.
Changes in government size positively correlate with the uncertainty and complexity of government’s organisational environment – as measured by the variance of GDP. Figure 4a shows the relationship between the uncertainty and complexity of government’s organisational environment (as measured by average variances in GDP) and changes in the size of government (as measured by average changes in total government
expenditure). For low-income, medium-income and high-income economies, more
output volatility corresponds roughly with more volatility in government expenditure during the 2000s. Such a correlation increases in strength for richer economies. Low- income economies exhibit a very weak pattern in the data while high-income economies show a relatively strong correlation between output volatility and the variance of
government expenditure.
0%
10%
20%
30%
40%
50%
0.0% 2.0% 4.0% 6.0% 8.0% 10.0% 12.0% 14.0%
ave. change in GDP
ave. change in expenditure Medium-Income
Lower-Income High-Income
Figure 4a: The Size of Government Probably Increases as the Government's Organisational Environment Becomes More Uncertain and Complex
The figure shows the relation between average rates of nominal GDP growth and average growth in government expenditure (in USD terms) from 1999 to 2009. Our averages represent simple arithematic means over the period. We use the World Bank's classification of countries by income-per-capita in assigning countries to income-groupings.
Source: World Development Indicators (2010).
Different types of governments adapt to changes in their organisational environment with different speeds. Figure 4b shows the correlation between changes in government expenditure and changes in GDP for the previous year, the current year and the
following year. We assume that differences between these correlations tell us about the government’s overall adaptive stance toward changes in the macroeconomy. For example, the figure shows the contemporary response (occurring in the same year) of changes in government spending to changes in GDP. Subtracting the difference between changes in government expenditure and changes in GDP between 2000 and 2008 gives a total “error” in government’s response to changes in output of roughly 36%.5
Depending on your view of the nature of change in government expenditure,
governments in high-income countries adjusted government sizes strategically while governments in medium-income countries adjusted contemporaneously. As shown in Figure 4b, between 1999 to 2003 changes in government expenditure relatively closely matched changes in output in both sets of countries. Only by 2004 did the “match”
between changes in government spending change significantly from changes in output.
By 2008, we observe changes in government spending again returning to a closer tracking of changes in output. Moreover, high-income countries’ governments tended to
5 We assume that policymakers will want to adjust government expenditure pro-cyclically with changes in GDP -- and by exactly the same percentage amount (in other words, unity represents the optimal elasticity of government expenditure with respect to GDP). Much empirical evidence suggests that policymakers instead adjust government expenditure counter-cyclically. In this case, the largest “errors” in the figure would best explain the government’s adaptive response to changes in its organisational environment. We use the figure to discuss the method of determining the government’s responsiveness to changes in its organisational environment – namely whether certain kinds of governments adaptive reactively, contemporaneously or strategically – rather than use the figure to pass judgments or make definitive conclusions about fiscal policy in these countries. We put the word “error” in quotes to emphasize that we take a positive rather than normative view of the data in this paper – seeking to describe the data rather than determine a best or optimal response.
match changes in expenditure (and thus probably government size) sooner and more closely with changes in output than medium-income countries’ governments. Figure 4b shows the annual differences between changes in government expenditure and changes in output – treating differences as an “error” (though such differences could reflect thoughtful policymaking in the presence of counter-cyclical organisational adaptation of organisational buffering against an excessively volatile organisational environment). For high-income countries, contemporaneous changes in output correlate less well with changes government spending than a similar correlation using lagged changes in output.
The difference between changes in output and government expenditure is almost twice as large if we assume that high-income country governments respond contemporaneous rather than strategically (changing government size before changes in output occur).6 For medium-income economies, however, the two approaches to government’s
organisational adaptation to changes in output yield roughly the same error. Using our measure of the “fit” of organisational response to changes in output, the figure shows that a model of contemporaneous response fit very well until about 2004 – whereas a model of strategic response fit less well. Thus, we have – for the purposes for labelling this set of countries in one category or the other – chosen to portray these countries governments’ organisational response as contemporaneous rather than strategic.
-10%
-5%
0%
5%
10%
15%
2000 2001 2002 2003 2004 2005 2006 2007 2008
"error" in expenditure change
Figure 4b: High Income Countries Adjust Expenditure Strategically while Medium-Income Countries Contemporaneously
High-income contemp response (36% "error")
Medium-income-
strategic response (42% "error")
High-Income
strategic response (19% "error")
Medium-income contemp response (43% "error") The figure shows the error in the adjustment of government expenditure to changes in GDP. The area under each line represents the difference between the change in government expenditure and GDP -- thus representing a type of "error" in government spending (assuming governments adapt pro-cyclically). We looked at three scenarios. We first substract the current year's change in GDP from the current year's change in government expenditure to investigate the extent to which government expenditure contemporaneously adjusts to changes in GDP. In the second set of calculations, we subtracted the current year's change in GDP from the following year's change in government expenditure to investigate the extent to which government expenditure changed reactively. In the third set of calculations, we subtracted the previous year's change in government expenditure from the current year's change in GDP in order to asses the extent of strategic change in government expenditure. We do not show reactive responses as they "fit" much less closely with changes in GDP.
Source: World Development Indicators (2010).
Government employment follows the same pattern of strategic adaptation to changes in the macroeconomic environment as expenditure does. Figure 5a shows three models of the “fit” of government employment to past, present, and future changes in output. When we compare the current year’s change in general government employment with changes in the current year’s GDP, we find a total “error” (as a defined previously) of 12.3%. In other words, using this measure results in about 12% difference between the sum of each year’s change in employment and output over the 9 year period.7 Assuming that high-
6 As described previously, we use the word “strategic” to describe changes in government expenditure occurring before changes in output. The lack of a response, or a counter-cyclical response may be more
“strategic” (as commonly understood in the public administration literature). We only use the word to describe changes in government spending in time and do not attach a value-judgment nor argue that strategic responses are necessarily superiour to other types of responses.
7 The graph starts at 2001 because the ILO report employment data only starting in 2000.
income economies governments’ adopted a reactive response to changes in output would result in a higher “error” of 13.5%. Yet, the reader should not interpret the lower “error”
as a better error. The higher error attached the lagged change in government employment could as well reflect organisational buffering – an organisational strategy aimed at insulating the organisation from pernicious changes in its external environment.8
-4.0%
-2.0%
0.0%
2.0%
4.0%
6.0%
2001 2002 2003 2004 2005 2006 2007 2008 2009
"error" in government employment
Contemporaneous Change Strategic Change Reactive Change
12.3% error for contemp adjustment
11.8% error for strategic adjustment
13.5% error for reactive adjustment
Figure 5a: High-Income Economies Exhibit Greater Strategic Organisational Change
The figure show s the "error" in the current year's, the previous year's and the follow ing year's adjustment in employment in the general government to changes in GDP for high-income countries. Controversially, w e describe series w ith the least difference betw een changes in GDP and changes in government employment and ignore the possibility of organisational buffering or counter-cycle employment practices. For example, for the high-income countries w e analyse, the sum of each year's difference betw een the current year's changes in GDP and previous year's change in government employment is low er than the sum of these differences using the current year's and the subsequent year's expenditure change. The three indicators show n in the figure represent w eighted averages, w here w eights come from each country's share in its income group's total employment.
Source: World Development Indicators (2010).
Medium-income countries’ government employment practices have responded much more sluggishly to changes in output. Figure 5b shows the annual difference between changes in GDP and changes in general government employment under 3 scenarios.
Assuming that government employment responds contemporaneously to changes in output results in a 38% total difference between the change of employment and changes in output between 2001 and 2009. Assuming that medium-income countries respond strategically – that employment adjusts before changes in GDP – results in a larger difference between overall changes in employment and output over the period than assuming that they adapt reactively.
8 Organisational theory does not provide specific predictions about the extent to which government organisations buffer against a highly volatile organisational environment. In theory, organisational complexity emerges to buffer the organisation from variation in the external environment up to a point.
After a certain size and age, very large and complex organisations learn to adapt to their external
environment (mostly out of necessity). Most academic commentators assume that buffering against largely un-diversifiable shocks represents one of the key functions of government – an assumption we do not make in this paper (as little large-scale empirical support for or against this hypothesis exists).
-10.0%
-8.0%
-6.0%
-4.0%
-2.0%
0.0%
2.0%
2001 2002 2003 2004 2005 2006 2007 2008 2009
Contemporaneous Change Strategic Change Reactive Change
Figure 5b: Organisational Response of Medium-Income Countries Much Sloppier in Responding to Changes in the External Policy Environment
41.9% error for strategic adjustment
37.3% error for reactive adjusmtnet 38.1% error for
contemp adjustment
The figure show s the "error" in the current year's, the previous year's and the follow ing year's adjustment in employment in the general government to changes in GDP for medium-income countries. Controversially, w e describe series w ith the least difference betw een changes in GDP and changes in government employment and ignore the possibility of organisational buffering or counter- cycle employment practices. For example, for the medium-income countries w e analyse, the sum of each year's difference betw een the current year's changes in GDP and subsequent year's change in government employment is low er than the sum of these differences using the current year's and the subsequent year's expenditure change. The three indicators show n in the figure represent w eighted averages, w here w eights come from each country's share in its income group's total employment.
Source: World Development Indicators (2010).
Changes in government real wages also support the conclusion that government size (and probably structure) in the high-income countries responds more to changes in the external macroeconomic environment than low-income economies’ government size and structures do. Changing real wages pull labour in, out, up and around government – serving as a useful proxy for larger structural changes in government.9 As shown in Figure 6a, the differences between changes in GDP and changes in real compensation paid to government employees appeared the greatest in the high-income economies.
Compensation for government employees (after adjusting for inflation) changed the least for low-income economies during the period – resulting in the largest differences
between changes in compensation and national output. For governments in countries of all income-levels, inflation-adjusted compensation for government employees fell during the period (as shown the constantly negative differences between changes in government employment compensation and changes in GDP). These trends contrast with the trends in nominal wages shown in Figure 2a – which showed large increases in nominal government wages.
9 In theory, changes in wages drive workers’ decision to accept government employment, promotions within government and serve as an important variable in workers’ decision to resign. As government departments emerge, expand, disappear – wages change (usually in practice through promotions or job category reassignments rather than explicit wage changes for the same job assignment). While government managers do not have the same right to engage in individual wage negotiations that their private sector counterparts have, they can greatly influence job reclassifications, promotions and reassignments which change the government worker’s wage.
-25%
-20%
-15%
-10%
-5%
0%
5%
2001 2002 2003 2004 2005 2006 2007 2008 2009
Figure 6a: High Income Economies Adapt Real Wages Most Closely to Changes in the External Environment
40% error for high incom e countries
98% error for m iddle incom e countries 117% error for low
incom e countries
The figure show s the "error" in the response of central government inflation-adjusted w ages to changes in GDP. Each indicator represents the w eighted average of the change in real (government) employee compensation (adjusted for local inflation) minus the change in real GDP for that year. We w eighted real employee compensations for each group of countries (given in the original data in local currency terms) by the total US dollar sum spent on government w ages for that country relative to the total spend for all governments in that country's income group.
Sources: IMF's Government Financial Statistics (2010) for (government) employee compensation and the World Bank's World Development Indicators (2010) for US dollar expenditure on government w ages.
Countries of all income levels “anticipated” falling output over the period with real wage compression. As shown in Figure 6b, the strategic scenario – where changes in the inflation-adjusted compensation of government employees precedes changes in output – fits more closely with changes in output than the other two scenarios. The overall difference or “error” centres on about 32% for high-income countries, 95% for medium- income countries and 112% for the few low-income countries for which the IMF provide data. In hindsight, reductions in government expenditure proved fortuitous in light of the sharp reductions in GDP (and thus in revenues) stemming from the 2009 global
economic crisis. Thus, to some extent, all countries governments engaged in “strategic”
adjustment (as we define the term strategic in this paper) of real wages (even if nominal wages increased during the same period).
20%
60%
100%
140%
Contempt Change Strategic Change Reactive Change
wage changes over or under-shooting GDP changes (smaller bars are "better")
High Income Middle Income Low Income
Figure 6b: Was Real Wage Compression in All Economies Prescient or Just Lucky?
The figure show s the "error" in w age adjustments over the period from 1999 to 2009 for low , middle and upper-income countries. The smaller bars for the upper-income countries mean that these countries changed public sector w ages more closely w ith changes in their countries' GDP. In order to calculate these errors, w e used the indicators show n in Figure 6a (w hich show w eighted averages of the real changes in public sector employees' compensation compared w ith changes in these countries' GDP). The contemporaneous change category show s the difference betw een changes in the current period's public sector w ages minus changes in GDP. The strategic change show s the difference betw een changes in real w ages for a period and the change in GDP in the follow ing period. The adaptive change category show s the difference betw een the previous period's real w age changes and the subsequent period's change in GDP. High- income countries significantly reduced real w ages throughout the 2000s, preparing them to adjust to the severe economic crisis of 2009.
Sources: IMF's Government Financial Statistics (2010) for (government) employee compensation and the World Bank's World Development Indicators (2010) for US dollar expenditure on government w ages.
Using unemployment rather than GDP volatility as the measure of the uncertainty and complexity of the government’s organisational environment produces much weaker correlations. Unemployment might serve as a better (or at least different) measure of the
uncertainty and complexity of the economic environment because unemployment represents a “bad” (which results according to popular expression from economic uncertainty and complexity) to which government should respond. Variance in output, on the other hand, does not represent a “good” or “bad” from a policy perspective.
Figure 7a shows the unemployment levels for all the countries for which we could obtain data -- compared with government size (as measured by government expenditure as a proportion of GDP). The data rather clearly show that government sizes do not respond to changes in overall unemployment – neither across country nor across time. We do not observe upward or downward sloping data; which we would expect for relationships across countries between these two variables. We neither observe gray dots “moving”
(as black dots) in any particular direction – as we might expect for a relationship
between these variables across time. As such, government does not act as an “employer of last resort,” shoring up unemployment during hard economic times. Government probably responds more to changing fundamentals in the macroeconomic environment rather than simply responding to domestic politics around employment (as we will discuss later in the literature review).
0 10 20 30 40
0 5 10 15 20 25 30 35 40 45 50
Governm ent Size (higher m eans larger governm ent) Unemployment (higher means more unemployed)
data in 2003 data in 2007
Figure 7a: No apparent pattern in the levels of government size and unemployment rates across countries nor with changes in those levels
between 2003 and 2007
The figure show s the relation betw een government expenditure (as a share of GDP) and unemployment (as a share of the labour force) in 2003 as compared w ith 2007 for high and medium-income countries. Each gray dot show s the combination of unemployment and government expenditure for a particular country in 2003. Each black dot represent that same pair of data in 2007.
Source: World Development Indicators (2010).
We do not observe a relationship between changes in government size and the
magnitude of asymmetric/sector-specific shocks. The data show – as shown in Figure 4a -- a relationship between the average size of shocks to a macroeconomy (which
presumably results in greater policymaker uncertainty in choosing correctly sizes and targeted policies) and the size of that country’s government. Yet, Figure 7b shows the relationship between the magnitude of asymmetric, sector-specific shocks – as measured by changes in industrial output relative to service-sector and agricultural sector output – and changes government size (as measured by expenditure). In the simple portrayal shown in Figure 7b, for economies of all income-levels, larger industrial sector shocks (relative to other sectors) do not correlate with changes in government size – as shown the circular clouds of dots in the figure.10
10 The correlation coefficients for each pair of data are all below 0.40 and not significantly different than zero. For high-income countries, the correlation coefficient equals 0.34, the coefficient for medium- income countries equals 0.16 and for low-income countries equals 0.22.
0%
25%
50%
-50% 0% 50% 100% 150%
average m agnitude of econom ic shocks average changes in expenditures and "shocks"
High Income Low Income Medium Income
Figure 7b: No Relation Between Average Changes in Expenditures and Changes in Economic Shocks
The data in the figure show the average changes in government expenditure compared w ith changes in the magnitude of economic shocks betw een 1999 and 2009. The index of economic shocks consists of subtracting the percent of industrial GDP from the percent of GDP in the service sector and dividing the resulting difference by agriculture's share in GDP. The average magnitude of economic shocks show n in the figure takes the arithematic average of changes in this index of economic shocks over the period 1999 to 2009 for each of the 96 countries show n in the figure. The average change in government expenditure show s the simple arithematic average change in dollar-valued government expenditure betw een the same period.
Source: World Development Indicators (2010).
The data also shows some validity for the resource-based view of organisational
structure – that government size responds more to tax and other resources available than to changes in government’s organisational environment? Figure 8 shows that such an explanation seems most plausible for medium-income countries – at least when looking at contemporaneous changes in government expenditure and revenue. Between 2000 and 2009, the sum of each year’s differences between low-income country governments’
expenditure and revenue resulted in an “error” (as we have previously defined such error) of 43%.11 Adjustment in high-income economies’ government expenditure showed an “error” of 47%. Medium-income country governments’ expenditure mismatch between expenditure and revenue over the period summed to 36%. The resource-based explanation of government organisation clearly provides some explanatory power – depending on the particular country and time.
11 Just like with our measure of adaptation to changes in government’s organisational environment, our measure of government’s “error” in responding to changes in resources only looks at the extent to which changes in government size contemporaneously adjusts to changes in revenues. Policymakers may wish to break the link between revenues and expenditure in any year in order to build up budget surpluses (in anticipation of future economic shocks), pay down previously acquired debts, or engage in fiscal policy to stimulate (or dis-stimulate) the macroeconomy. Given this wide range of organisational objectives, we only report the positive aspects of organisational adaptation -- ignoring the normative aspects (dealing with the desirability and/or optimality) of such changes.
-20%
-10%
0%
10%
20%
2000 2001 2002 2003 2004 2005 2006 2007
Figure 8: Resource-Based View of Government Organisation Most Valid for Middle Income Countries in the Short-Run
Medium-Income countries
The figure shows the difference between changes in government expenditure and revenue for each of the three categories of countries shown. The difference between changes in contemporaneous government expenditure and revenue (with changes occuring in the same year) is lowest across the entire period for medium-income countries -- which we interpret as most strongly supporting the resource-based view of government (that government revenue determines expenditure rather than other considerations). To derive these indicators, we took the weighted average of changes in US dollar valued expenditure and revenue. Weights for changes in government revenues consisted of the country's 2004 government revenue (in US dollars) as a percent of the total revenue (also expressed in US dollars) for that country's income group. Weights for government expenditure consisted of that country's 2004 expenditure (expressed in US dollars) as a percent of that country's income group's total expenditure (again expressed in US dollars). Source: World Development Indicators (2010).
High-income countries Low-income countries
Different countries’ governments adapt their organisational sizes at different speeds in response to changes in their organisational environments. Figure 9 shows the best fitting (possessing the least amount of “error”) adaptive orientation for various countries’
government sizes among strategic, contemporaneous, reactive and resource-based models of organisational adaptation to changes in the macroeconomic environment (as measured by the change in industrial GDP relative to other sectors). In general, the changes in the size of governments like those of the USA, China and Finland correlated more closely with changes in the sectoral distribution of output before such changes in output occurred. Changes in government sizes for countries like India, Australia, Kazakhstan and Argentina tended to correlate with changes in sector output as such changes in sectoral output occurred. Changes in government size for countries like Russia, Algeria, Germany and the UK tended to correlate with changes in industrial output (relative to other sectors) only after such changes in industrial output occurred.
Finally, for countries like Canada, Iran, and Sweden, changes in government sizes correlated most closely with the revenue these governments had at their disposal in any given year.
Each theory of government organisation provides a partial explanation for these data.
Even the most die-hard critics of the contingency theory of government organisation must acknowledge that government size should respond (at least in part) to changes in the macroeconomic environment. Fiscal policy (namely government expenditure on goods, staff and assets like office desks) – by law if not by practice in many countries – smoothes out the effects of general and asymmetric macroeconomic shocks (which would be seen in relatively low correlations between changes in government sizes and macroeconomic changes in some countries). Critics of the resource-based theorists can not argue that governments can not expand beyond their means in the long-run (namely their revenue and borrowing power). Critics of rational-choice theorists can not argue that government can anticipate many kinds of shocks – rising grain or oil prices,
demographic changes and so forth. Some of the “strategic” organisational adaptation we observe in the data probably does reflect actual strategic policymaking. Yet, some of the
“reactive” organisational adaptation we observe in the data may reflect rational organisational buffering or anti-cyclical spending.
A Background on the Structure of Government
Changes in the size and composition of government expenditure must translate into changes we typically think of as the “structure of government.” When an organisational theorist thinks about the structure of structure, concepts like the number of
organisational units or agencies, the number of staff in a department, the number administrative or budgetary departments in a division or directorate come to mind. No public data exist on these classical organisational features. Yet, we can infer changes in the structure of government by looking at changes in the composition of government expenditure. Rather than define a formal model, we will present a very simple thought- experiment.
A simple thought-experiment shows the relation between the distribution of government expenditure and more classical notions of the “structure of government.” Imagine you
hold a job in a government ministry (or department) in the late 1990s -- when government departments started using IT-technologies extensively (though the
introduction of an anti-terrorism programme or any other policy initiative will do). The ministry of finance allocates money in order for your government department to install and use a couple of computers. Your minister might give an IT specialist(s) an office and maybe even create a separate small organisational unit for these IT specialists. Now image – as shown in Figure 10a – changes in IT technology (and thus large changes in the country’s IT production or imports) cause the ministry of finance to allocate $20 million more in resources for the computerisation of your department. Your own minister or boss must spend the money somewhere. Your boss hires more people, purchases more equipment and so forth. After a point, the original IT manager can not cope with all the new staff. He or she can not manage unlimited amounts of staff and assets. Only three “structural” solutions exist – make the IT unit bigger (into a division or department), split it up by functions and scatter those functions around your ministry, or send it partially (or completely) outside your department. No other organisational responses exist. We can not deduce – only by looking at changes in resources – how organisational structure changed. But we can be fairly certain – particularly when we collect data about the large numbers of governments and their changing composition of expenditure – that the “structure of government” (as a classical organisational theorist would understand the term) changes as the composition of government expenditure changes.
organisational unit obtains twice the resources…
Make a new department or division (or both)….
…or establish new structures outside of our existing department (or make cross-agency one)
Resources “entering” the organisation must go somewhere…. and limits on spans of control, office space, and group-dynamics prevent unlimited expansion without more fundamental changes in the
“structure” of the organisation (though not true if only storable financial resources come in).
Figure 10a: The Intuition Behind the Link Between Changes in the Composition of Resources and Changes in the Structure of Government
We can infer that the “structure of government” probably differs between countries.
Figure 10b provides a comparison of the allocation of expenditure across functional categories between high-income and medium-income countries (for ease of exposition, though we could show differences for each country). High-income and medium-income countries tend to allocate the same proportion of expenditure on general government services (about 22%), education (about 10%) and health (about 9%) – though with very significant variation between countries (which we do not show). Yet, even average levels differ greatly between high-income and medium-income countries in defence and public order (with a 3% difference) and economic affairs and social protection (with about an 8% to 10% difference). We know from anecdotal evidence and just plain
